🎁 New Sign up free, 10 calls on us. Up to $1, no card needed.

Fun-ASR Flash vs Seed ASR

vs

Which one, when — curated verdict, not a benchmark table

Per-minute rates are within a rounding error ($0.0021 against $0.002), which puts the choice entirely on capability: seed-asr-bigmodel diarizes on its audio-file API (documented best at 10 speakers or fewer), returns sentence and word timestamps and accepts up to 5 hours per file, while fun-asr-flash is synchronous only, capped at 5 minutes and 2GB, with no diarization but with context injection for domain terms. Pick fun-asr-flash for fast short-clip turnaround, seed-asr-bigmodel for anything long or multi-speaker.

Pricing

Fun-ASR Flash Seed ASR Δ
Per audio minute $0.0021 $0.002

Rates from the live catalog at build time; each model page carries the current card.

Where they sit — price per audio minute across all 11 speech-to-text models on this billing unit (log scale)

Fun-ASR Flash · $0.0021 Seed ASR · $0.002
$0.002 · Fun-ASR Realtime $0.016 · Chirp 2

Capabilities

Fun-ASR Flash Seed ASR
Speaker diarization no yes
Streaming yes yes
Timestamps yes

Specs

Fun-ASR Flash Seed ASR
Input modalities audio audio
Output modalities text text
Released 2026-06
Limits

Synchronous fast recognition, <=5 min / <=2GB per audio

context injection for domain terms

30+ languages

no diarization

Async audio-file mode: <512MB, <5 hours, OPUS/WAV/MP3/SPX/OGG/AMR/AAC/M4A (raw PCM also accepted), results returned within 3 hours and retained 7 days

speaker diarization via enable_speaker_info (audio-file API only, best with <=10 speakers, no diarization on the streaming API)

sentence + word segmentation with start_time/end_time via show_utterances

language identification via enable_lid

hotwords/context up to 800 tokens and 20 rounds

per-call billing

Languages Multilingual with dialects, the same 30-language list as the Fun-ASR main versions: Chinese (Mandarin, Cantonese, Wu, Hokkien, Hakka, Gan, Xiang, Jin plus regional accents), English, Japanese, Korean, Vietnamese, Thai, Indonesian, Malay, Filipino, Hindi, Arabic, French, German, Spanish, Portuguese, Russian, Italian, Dutch, Swedish, Danish, Finnish, Norwegian, Greek, Polish, Czech, Hungarian, Romanian, Bulgarian, Croatian, Slovak

With `language` empty the model covers Mandarin, English, Cantonese, Shanghainese, Minnan, Sichuan and Shaanxi dialects

39 language keys can be pinned explicitly (en-US, zh-CN, yue-CN, ja-JP, ko-KR, id-ID, es-MX, pt-BR, de-DE, fr-FR, fil-PH, ms-MY, th-TH, ar-SA, it-IT, bn-BD, el-GR, nl-NL, ru-RU, tr-TR, vi-VN, pl-PL, ro-RO, uk-UA, az-AZ, bg-BG, cs-CZ, da-DK, fi-FI, hi-IN, hu-HU, kk-KZ, km-KH, my-MM, no-NO, pa-PK, sv-SE, sw-KE, ur-PK), plus optional auto-detection (enable_auto_lang)

Specs are transcribed from each vendor’s documentation; a row a vendor does not publish is left out rather than inferred. Full sources: Fun-ASR Flash · Seed ASR

Switch between them with one line

Both ids are in every tab below — the highlighted pair of lines is the only edit. Same endpoint, same key, same request shape.

from openai import OpenAI

client = OpenAI(
    base_url="https://synthorai.io/v1",
    api_key="sk-syn-...",
)

resp = client.audio.transcriptions.create(
    model="fun-asr-flash",
    # model="seed-asr-bigmodel",  # uncomment this line, comment the one above
    file=open("meeting.mp3", "rb"),
    language="en",
)
print(resp.text)

Get an API key →

FAQ

Which is cheaper, Fun-ASR Flash or Seed ASR?

Seed ASR is cheaper on per audio minute ($0.002 vs $0.0021, 1.0× apart). Other rows may point the other way — the table above carries the full card, and real cost depends on your mix.

Can I A/B test Fun-ASR Flash against Seed ASR without two integrations?

Yes. Both are served through the same OpenAI-compatible endpoint with one API key — switching is a one-line model-string change, so you can route a fraction of traffic to each and compare bills directly.

Do Fun-ASR Flash and Seed ASR support speaker diarization?

The capability table above answers this per model, straight from each vendor’s documentation — diarization, streaming and timestamps are listed separately because models differ on all three.

Related comparisons

From our measured studies